J. MATER. CHEM., 1991, 1(3), 361-364 361 Evolution of Structural Changes during Flash Calcination of Kaolinite A 29Siand *'AI Nuclear Magnetic Resonance Spectroscopy Study Robert C. T. Slade"" and Thomas W. Davied a Department of Chemistry, Universiiy of Exeter, Exeter EX4 4QD, UK Department of Chemical Engineering, University ofExeter, Exeter EX4 4QF; UK Kinetically frozen samples of flash calcined kaolinite (rapidly heated to 1000 "C,maintained at that temperature for a variable residence time and then rapidly cooled) have been produced in a laboratory calciner. Time resolution of the structural changes occurring has been achieved by following 27AI (78.15 MHz) and %i (59.58 MHz) magic angle spinning nuclear magnetic resonance (MAS NMR) spectra as a function of residence time.29Si spectra can be deconvoluted into a minimum of four Gaussian components with shifts ranging from mullite-like (-90 ppm) to Q4 (-110 ppm) Si environments. 27AI spectra show peaks for four-co-ordinate and six- co-ordinate Al. The derived picture of flash calcination is progressive transformation of kaolinite to a single product which undergoes little further chemical reaction during its short time in the calciner. Keywords: Flash calcine; Kaolinite; Magic angle spinning nuclear magnetic resonance spectroscopy; Deh ydroxyla tion Kaolinite [china clay, A12Si205(OH),, sometimes written as A1203.2Si02 *2H20] is a raw material of considerable indus- trial significance on an international scale.' The kaolinite structure is built by stacking lamellae composed of a pair of silica and alumina sheets.The silica sheets contain vertex- shared SiO, tetrahedra, while the alumina sheets contain edge-shared A106 octahedra. Kaolinite calcination (dehydroxylation) to form the ther- mally stable compound metakaolin (A12Si207, A1203*2Si02) is an important step in the manufacture of clay products. Reorganisation of the structure during dehydroxylation is dominated by forces within the alumina sheets. Three of the four hydroxyl groups associated with the alumina sheets lie in the interlamellar space between successive sheet pairs, while the fourth is intralamellar (between the silica and alumina sheets). Dehydroxylation is a rate process and calcines can be kinetically frozen at various stages of structural reorganis- ation.In the absence of structural collapse, an idealised kaolinite would lose 13.95% of its mass on complete dehy- droxylation and its density would drop from 2.64 to 2.27 g cm-3.2 Two very different methods for dehydroxylation are used, soak calcination and flash calcination. Soak Calcination In industrial soak calcination the dehydroxylation is achieved by holding the clay at a sufficiently high temperature (600< T/ "C<1000) for a sufficient length of time (ca.1 h) in an oil- or gas-fired furnace. The rate at which the clay is brought to the calcination temperature is low and is not used as a process variable; the resulting metakaolin has a density of ca. 2.74 g cm-3.3 The properties of metakaolin formed by soak calcination have been the subject of extensive studies (see e.g.ref. 4) and are a useful benchmark against which the properties of flash calcines (see below) can be compared. Overheating kaolinite results in the formation of mullite (A16Si2013)and cristobalite (SO2), generally considered unde- sirable (as the abrasiveness of the calcine is increased). Much effort has been devoted to studies of the physical and chemical changes associated with this reaction e.g. ref. 5-9. The use of analytical techniques such as MAS NMR has revealed more detail about structural changes accompanying the dehydroxyl- ation step e.g. ref. 10. Flash Calcination If kaolinite particles are heated at such a speed that the steam released within them is generated faster than it can escape by diffusion, structural disruption is likely.Such structural dis- ruption may endow the resulting calcine with desirable or interesting properties, such as internal voids. ''The diameters of such voids are comparable to the wavelengths for visible light, therefore producing light scattering and imparting opac- ity to the material (then usable as an effective paper covering). Construction at Exeter of a furnace to allow kaolinite particles to be subjected to thermal histories comparable to those in industrial flash calciners has been described elsewhere." In industrial flash calcination, cold powdered clay is passed through a gas or oil flame and then quenched by injection of cold air.The laboratory simulation of this process involves plunging a stream of clay particles into a co-flowing stream of hot He(g) (which is in downward laminar flow) in a vertical electrically heated reaction tube. Flash calcines produced in this way have quite different properties from corresponding soak calcine^."^'^-'^ We have previously demonstrated the utility of MAS NMR techniques in probing the structural consequences of introduc- ing water vapour into the calciner atmosphere (variation of H20 content of the He carrier gas) in production of flash calcines of similar densities.16 We now report the use of NMR absorption spectra to gain insight into the evolution of structural changes during the flash calcination of kaolinite, with time resolution being achieved via variation of the residence time at the reaction temperature.Experimental The kaolinite feedstock was commercial grade SPS clay (English China Clays, St. Austell, Cornwall). 90% of the powder was <2 pm particle size. XRF analysis gave Si02 46.2%, Al2O3 38.7%, Fe203 0.56%, Ti02 0.09%,CaO 0.20%, MgO 0.20%, K20 1.01%, Na20 0.07% and the loss on ignition ('H20 content') was 13.14%. Partially dehydroxylated flash calcines were prepared using the laminar-flow furnace described previously. ' This allows powdered clay to be heated rapidly (in a few ms) from room temperature to a controlled temperature (ca. 1000 "C) and then cooled back to room temperature by quenching with cold N2(g). The residence time at the reaction temperature is controllable and was the only process variable explored in this study.Residence times (at 1000 "C in dry He) were varied in the range 0.2-0.8 s. Fig. 1 presents the results of density determinations (p, by water displacement according to BS 190-304) and degree of dehydroxylation (a, by subsequent thermogravimetric dehy- droxylation to completion on a Stanton STA-780 instrument) for the calcines produced. X-ray powder diffraction patterns (Ni-filtered Cu-Ka radi-ation) were recorded using a computer-controlled Philips PW 1050 goniometer incorporating accumulation of multiple scans. In calcines with residence times up to 0.3 s little change from the kaolinite pattern was seen. At longer residence times, scattering from 'amorphous' (transformed) material became dominant, with progressive loss of intensity from a residual kaolinite component.At no stage was any other phase (e.g. mullite or cristobalite) observed in the diffraction pattern. Proton-decoupled high-resolution MAS NMR spectra for 27Al (78.15 MHz) and 29Si (59.58 MHz) were recorded at ambient temperature using a Varian VXR300 spectrometer. For 29Si spin rates of ca. 5 kHz and n/2 radiofrequency pulses were used. For 27Al a high-spin-rate (Doty) probe was used with spin rates of 9-10 kHz and employing 46 radiofrequency pulses. A small signal due to A1 in the probe was subtracted from the recorded 27Al spectra. Relaxation delays were varied to ensure the absence of saturation effects from reported 2.6 6 I 1 I I I I \ 2.55- \ 2.5- \ 5 2.45-2.4-(1 I a Q \ \ \Q \ \ t 2.35- \ / t 2.3/ 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 residence time/s 80 o/ / czj 404 /I / I '0 204 4' Fig.1 Effect of particle residence time on the density (p) and degree of dehydroxylation (a) of the resulting calcine J. MATER. CHEM., 1991, VOL. 1 spectra. In the case of 29Si spectra necessary relaxation delays depended on the calcine (e.g. 60s for kaolinite and short residence-time calcines, 2.0 s for calcines at the longest resi- dence times), while delays of 1.0 s were employed in obtaining 27Al spectra. Spectra are referenced to tetramethylsilane (2gSi) and Al(H20)z+(aq) (27Al). Results Fig.2 shows the 29Si spectrum as a function of residence time. At residence times longer than 0.3 s the kaolinite contribution (line at ca. 91.5 ppm) decreases progressively, indicating (in accord with the X-ray studies above) progressively less unchanged material, and the (broad) absorption due to trans- formed material appears. As with soak calcines," the latter absorption extends from mullite-like (ca. -85 ppm) through to Q" (ca. -1 10 ppm) Si chemical shifts. With increasing residence time the maximum in the broad absorption appears to move progressively upfield, but this is consequential on the observation of superimposed spectra from kaolinite and transformed material (see below). Fig. 3 shows the 27Al spectrum as a function of residence time.No significant or systematic variations in shifts with residence time were seen. Peaks assignable to octahedral A1 were observed at -3.1 k0.3 ppm, with those assignable to tetrahedral A1 at 55.5 & 0.4 ppm. At short residence times a dominant peak assignable to octahedral A1 is observed. At residence times longer than 0.3 s a peak assignable to tetra- hedral A1 and a broad underlying signal (-50-75 ppm) arising from A1 in distorted environments appear, both increasing in intensity with increasing residence time. Discussion Relationship to Spectra of Soak Calcines A number of high-resolution NMR studies of soak calcines have been reported. "7 17-22 II -80 -100 -120 c, -80 -100 -120 -80 -100 -120 6 (PPm) 6 (PPm) Fig.2 29Si MAS NMR spectra of flash-calcined kaolinite as a function of residence time 7 at lo00 "C in He z/s: (a)0.2, (b) 0.3, (c) 0.4, (d) 0.5, (e)0.6, (f)0.7, (g)0.8 J. MATER. CHEM., 1991, VOL. 1 ------L 100 ' 76 100 A -1 00 ----qA%100 100 0 -100 c_?v/ 8 (PPm) 100 0 -100 6 (PPm) Fig. 3 "A1 MAS NMR spectra of flash-calcined kaolinite as a function of residence time z at 1000 "C in He z/s: (a) 0.2, (b) 0.3, (c) 0.4, (d) 0.5, (e) 0.6, (f) 0.7, (g) 0.8 29Si spectra of soak calcines are broad (as in this study of flash calcines) with features extending from Si shifts character- istic of mullite-like environments through to Q" Si (superscript denotes number of Si-0-Si bridges per tetrahedral Si, Si in kaolinite itself is Q3). Spectra are consistent with the presence of a range of environments Q", each of which includes sites with geometries grossly distorted from those in parent kaolinite.Lambert et ~1.~'deconvoluted spectra into the minimum number of overlapping Gaussians, this giving suggestive information concerning abundances of different Q" types as a function of (soak) calcination conditions. Higher numbers of Gaussians have been mooted,21.22 but the choice then becomes somewhat arbitrary. Discussions of 27Al (nuclear spin I=3) spectra commonly neglect field-dependent second-order quadrupole shifts (aqs)and line-broadenings (the true field-independent chemical shift ~cs=~c~I&where aCG is the centre of gravity of the observed lineshape22).Consideration of these effects would be problematic in these systems. 27Al spectra of soak calcines show peaks at ca. 0 ppm (octahedral Al), at ca. 55 ppm (tetra- hedral Al) and at ca. 30 ppm. The peak at 30 ppm is now believed to arise from pentaco-ordinated Al, this assignment being unambiguous in the case of andal~site~~ and also being made in the case of a variety of thermally/hydrothermally treated alumino~ilicates.~~ No spectral feature unambiguously assignable to pentaco-ordinated A1 was observed in this study (of flash calcines). The picture of soak calcination that has emerged from such studies is as follows: (1) transformation of kaolinite to meta- kaolin is accompanied by disorder and distortions in Si environments; (2) treatment to higher temperatures leads to segregation of regions of 4" Si, mullite and a spinel phase (Sanz et a1.,I9 Lambert et aL2' and Rocha and Klinowski" are contradictory as to whether this could be y-A1203).29SiSpectra of Flash Calcines The broad absorption characteristic of transformed material is indicative (as in the case of soak calcines) of a range of (~3,)~ Gaussian components characterised by their chemical shifts distorted Si environments of differing Q". The spectra in Fig. 2, which contain the broad absorption (residence time >0.3 s) can each be deconvoluted into a minimum of four Gaussian components, the results of this deconvolution being given in Table 1. Fig. 4 shows the resulting fit and contributory Gaussians for the calcine with residence time 0.4s.The Gaussians are assignable (in order of increasingly negative 6,) to mullite-like, residual kaolinite-like (in untransformed material and possibly within the product particles also), metakaolin and Q" Si environments. In accord with this assignment (i) the kaolinite peak is by far the narrowest and its intensity decreases with increasing residence time (as would be predicted from the X-ray studies above) and (ii) the relative intensities of the mullite-type and Q" peaks are approximately in the ratio 1 :2 (which would be that observed in the event of phase segregation). Table 1 Deconvolution of 29Si spectra for flash calcines into four widths (FWHM) and intensities residence time/s 6, (ppm) FWHM/Hz intensity (YO) Gaussian 1 (mullite-like) 0.4 -90.9 825 18 0.5 -91.0 754 20 0.6 -90.7 762 13 0.7 -88.9 852 18 0.8 -88.8 798 15 0.4 Gaussian 2 (kaolinite) -93.1 122 11 0.5 -92.5 121 9 0.6 -93.4 142 6 0.7 -92.6 185 5 0.8 -92.6 202 3 0.4 Gaussian 3 (metakaolin) -98.6 768 34 0.5 -99.1 806 36 0.6 -99.2 768 40 0.7 -100.2 794 45 0.8 -100.2 823 43 0.4 Gaussian 4 (Q4, cristobalite-like) -109.5 909 36 0.5 -109.6 856 35 0.6 -110.1 830 41 0.7 -109.6 765 33 0.8 -1 10.0 862 38 -40.2 6 (PPm) -166.2 Fig.4 Deconvolution of the 29Si MAS NMR spectrum of flash- calcined kaolinite with a residence time r=0.5 s (at lo00 "Cin He) into four contributing Gaussian components (see text).The experimen- tal data, the final fit (smooth bold line) and their difference are shown The chemical shifts of the peaks associated with changed Si environments remain approximately constant with increas- ing residence time and degree of conversion. It follows that the apparent upfield shift of the maximum in the broad absorption with increasing residence time is a consequence of the decrease in kaolinite-like sites present, rather than of a change in the transformed material. The proportion of Si in new sites that are assigned to metakaolin-like environments appears to increase slightly with increasing residence time (39,40, 43, 46 and 45% of new sites at 0.4,0.5,0.6,0.7 and 0.8 s, respectively).This could be linked with the increasing density of the calcine evident in Fig, 1. *'A1 Spectra of Flash Calcines Quantitative deconvolution of 27Al spectra is not particularly meaningful because of (i) the contribution from A1 in highly distorted environments being quadrupolar broadened beyond observation (there is a corresponding loss in intensity) and (ii) second-order quadrupolar shifts, linewidths, unknown lineshapes and overlapping contributions. The lack of a spectral feature assignable unambiguously to pentaco-ordinate A1 in this study is a significant difference from the spectra of metakaolin produced by soak calci- nation.20g22A broadened contribution to the underlying signal (-50-75 ppm) from distorted pentaco-ordinate A1 cannot be ruled out however.It should be emphasised that the use of high spin rates is essential in this work. Use of a lower spin rate (such as the more usual ca. 3 kHz at the Larmor frequency in this study) would result in a spinning sideband close to the anticipated location of a signal from pentaco-ordinate A1 and consequent difficulty in deriving any conclusion as to the occurrence of such a feature. The peak shown by detectable A1 atoms at 56ppm is attributable to tetrahedral Al. This supports the formation of some three-dimensionally connected aluminosilicate-like (tetrahedral Al) regions during dehydroxylation. The suppo- sition that the silica and alumina sheets are preserved during dehydr~xylation'~is therefore too simplistic, The proportion of observed A1 in tetrahedral sites increases with increasing residence time, this variation correlating with the decreasing proportion of untransformed material evident in X-ray pat- terns and 29Si spectra.The increasing relative intensity assign- able to tetrahedral A1 arises from the increasing proportion of transformed material. Conclusion The observed variations in densities, degrees of dehydroxyl- ation and NMR spectra suggest the following temporal evol- ution of structural changes during flash calcination: (1) At short times (<0.3 s) most of the kaolinite is largely unchanged (as evident from NMR spectra), while the bulk density decreases (and degree of dehydroxylation increases) owing to transformation of surface layers.(2) At longer times (>0.3 s) dehydroxylation within the particles commences, with accompanying 'bubbling' (internal void formation) and density loss. At this stage, Si within the material migrates increasingly to three non-kaolinite environ- ments (termed mullite-like, metakaolin-like and Q").Some A1 migrates to tetrahedral sites. (3) The relative proportion of Si in transformed regions that is assigned as metakaolin-like environments increases slightly at the longest residence times. (4) As residence time increases and the amount of untrans- formed material decreases, the relative proportion of A1 in tetrahedral environments increases (correlating with the pro- portion of product). J. MATER. CHEM., 1991, VOL. 1 The derived composite picture of flash calcination is that of a material transforming increasingly from kaolinite to a single product.Once formed, the product appears to undergo little further chemical reaction during its short time in the calciner. Prolonged (soak) thermal treatment of flash calcines pro- duces materials comparable (e.g.densities) to those from soak calcination. The transformations towards those products are much slower than the initial step giving the flash calcine. It should be noted, however, that slightly increasing densities at the longest residence times in this work could be evidence for shrinkage of internal voids (by atom migrations in the trans- formed material at the calciner temperature), this process being accompanied by a slightly increasing proportion of metakaolin-like Si environments.The detailed interpretation of 29Si and 27Al NMR spectra for calcines is likely to be highly complex. For soak calcines detailed models of thermal transformations of structures have been pr~posed,'~.'~ but understanding remains incomplete. For the flash calcines produced in this study, further NMR investigations are in hand using cross-polarisation ('H-29Si CP-MAS) and nutation (27Al)techniques. We thank SERC for supporting this study under grant GR/E 81999 and for access to the National Solid State NMR Service (University of Durham). We thank the staff of that service for recording high-resolution spectra and for subsequent computations. References 1 W.D. Keller, Geology Today, 1985, 109. 2 G. W. Brindley and M. Nakahira, J. Am. Ceram. SOC., 1959, 42, 31 1. 3 R. E. Grim, Clay Mineralogy, McGraw-Hill, New York, 2nd edn., 1968. 4 M. C. Gastuche, F. Toussaint, J. J. Fripiat, R. Touilleaux and M. van Meersche, Clay Miner. Bull., 1963, 5, 227. 5 G. W. Brindley and G. L. Millhollen, Science, 1966, 152, 1385. 6 G. W. Brindley, personal communication, 1984. 7 C. Otero-Arean, M. Letellier, B. C. Gerstein and J. J. Fripiat, in Proc. Znt. Clay Con5 1981-BoIogna, ed. H. Van Olphen and F. Veniale, Elsevier, Amsterdam, 1982, p. 73. 8 G. W. Brindley, J. H. Sharp, J. H. Patterson and B. N. Narahari Achar, Am. Miner., 1967, 52, 201. 9 J. M. Criado, A. Ortega, C. Real and E. Tores de Torres, Clay Miner., 1984, 19, 653.10 R. H. Meinhold, K. J. D. Mackenzie and I. W. M. Brown, J. Mater. Sci. Lett., 1985, 4, 163. 11 D. Bridson, T. W. Davies and D. P. Harrison, Clays Clay Miner., 1985, 33, 258. 12 T. W. Davies, High Temp. Technol., 1984, 2, 141. 13 T. W. Davies, Chem. Eng. Res. Des., 1985, 63, 82. 14 T. W. Davies and R. M. Hooper, J. Mater. Sci. Lett., 1985, 4, 39. 15 T. W. Davies, J. Mater. Sci. Lett., 1986, 5, 186. 16 R. C. T. Slade and T. W. Davies, Colloids SurJ, 1989, 36, 119. 17 K. J. D. Mackenzie, I. W. M. Brown, R. H. Meinhold and M. E. Bowden, J. Am. Ceram. SOC., 1985, 68, 293. 18 I. W. M. Brown, K. J. D. Mackenzie, M. E. Bowden and R. H. Meinhold, J. Am. Ceram. SOC., 1985, 68, 298. 19 J. Sanz, A. Madani, J. M. Serratosa, J. S. Moya and S. ha, J. Am. Ceram. SOC., 1988, 71, 418. 20 J. F. Lambert, W. S. Millman and J. J. Fripiat, J. Am. Chem. SOC.,1989, 111, 3517. 21 J. Rocha and J. Klinowski, Angew. Chem., Znt. Ed. Engl., 1990, 29, 553. 22 J. Rocha and J. Klinowski, Phys. Chem. Miner., 1990, 17, 179. 23 E. Lipmaa, A. Samoson and M. Magi, J. Am. Chem. SOC., 1986, 108, 1730. 24 J. Gilson, G. C. Edwards, A. W. Peters, K. Rajagopalan, R. F. Wormsbecher, T. G. Roberie and M. P. Shattock, J. Chem. SOC., Chem. Commun., 1987,9 1. Paper 01049 17D; Received 1st November, 1990